llm-app

Pathway LLM App provides ready-to-deploy AI pipeline templates for building high-accuracy RAG (Retrieval-Augmented Generation) and enterprise search applications at scale. The templates connect to live data sources including local file systems, Google Drive, SharePoint, S3, Kafka, PostgreSQL, and real-time data APIs, automatically syncing additions, deletions, and updates. Application templates include: Question-Answering RAG App for document Q&A, Live Document Indexing as a vector store service, Multimodal RAG with GPT-4o for extracting tables and charts from PDFs, an Unstructured-to-SQL pipeline that converts financial documents into structured SQL queries, Adaptive RAG to reduce token costs, Private RAG using Mistral and Ollama for fully local deployment, Slides AI Search for PowerPoint and PDF indexing, and Video RAG with TwelveLabs for video content search. The framework runs as Docker containers with an HTTP API and optional Streamlit UI. It eliminates the need for separate vector databases, caches, or API frameworks by providing built-in vector indexing (via usearch) and hybrid full-text search (via Tantivy), all processing in-memory with caching. Templates scale to millions of pages and can be deployed on GCP, AWS, Azure, Render, or on-premises. Each template includes a README with setup instructions, and additional code templates are available on the Pathway website.